brigade-exporter

brigade-exporter

brigade-exporter is a Prometheus metrics exporter for Brigade.

This exporter is designed to be run along with a brigade installation, if you have multiple brigades you will have multiple brigade-exporters, one per brigade installation. This follows the philosophy of prometheus exporters of one exporter per app instance.

Run

There is already a docker image ready to run the exporter in quay.io/slok/brigade-exporter. It has different options to run.

Run outside the cluster

If you want to test the exporter outside the cluster in a brigade installation, you can use --development flag. You will need kubectl configuration and the context set pointing to the desired cluster.

docker run --rm \
    -p 9480:9480 \
    -v ${HOME}/.kube:/root/.kube:ro \
    quay.io/slok/brigade-exporter:latest \
    --debug \
    --development \
    --namespace ${MY_BRIGADE_NAMESPACE}

go to http://127.0.0.1:9480/metrics

Grafana dashboard

• Brigade dashboard: A grafana dashboard for brigade.

Deployment

TODO

RBAC

TODO

Metrics

Exporter metrics

Project metrics

Build metrics

Job metrics

Disabling metrics

You can disable metrics using flags.

• --disable-project-collector: Disables all the metrics of projects.
• --disable-build-collector: Disables all the metircs of builds.
• --disable-job-collector: Disables all the jobs metrics. If you have lots of jobs, this could improve the gathering and storage of metrics.

Build from source

You can build your own brigade-exporter from source using:

make build-binary

to build the binary or

make build-image

to build the image.

Development

Run in fake mode

If you are developing, the exporter can fake a brigade installation and return fake data using --fake flag.

Run the stack with a configured Prometheus

If you want to run a local exporter+prometheus stack run.

make stack

And you will have a prometheus on http://127.0.0.1:9090 that will scrape a faked brigade-exporter.

Query examples

% of running builds per provider.

sum(
    brigade_build_status{status="Running"} * on(id) group_right brigade_build_info
) by (provider)
/ on() group_left
sum(
    brigade_build_status{status="Running"})
* 100

Get the jobs and their states of a build

brigade_job_info{build_id="build-xxxx"}
*on(id) group_right brigade_job_status

Get how long the jobs have been in pending state before started to run.

(brigade_job_start_time_seconds > 0) - (brigade_job_create_time_seconds > 0)

Get the top 10 project builds duration by event and provider (in the last 30m)

topk(10,
  avg(
    max_over_time(brigade_build_duration_seconds[30m])
      * on(id) group_right brigade_build_info
    * on(project_id) group_left(name)
      label_replace(brigade_project_info , "project_id", "$1", "id", "(.*)")
  ) by(name, provider, event_type))

Average job duration seconds per project Note This is an extreme example of how you owuld scalate IDs in metrics. This is not recommended.

avg(
label_replace(
    label_replace(
      avg(
        (brigade_job_duration_seconds > 0) * on(id) group_right brigade_job_info
      ) by (build_id)
    , "id", "$1", "build_id", "(.*)")
    * on(id) group_right brigade_build_info
, "id", "$1", "project_id", "(.*)")
) by (id)
* on(id) group_right brigade_project_info